Projective Structure, $\Widetilde {\Mathrm {SL}}(3,{\Mathbb R}) $ and the Symplectic Dirac Operator

Home , Real projective space

arXiv:1604.04376v1 [math.DG] 15 Apr 2016 n nepeainwsntcera httm.Hr efidana a find we Here time. that alre at algebra structure clear Lie this not was on symmetries interpretation information differential and Some of two. question dimension the real namely work, rece our the for from comes development [1]. this projec operator, for Dirac of symplectic geometry inspiration the An of) covering two. double (the of framework eciearaiaino ypetcsio ed n h sy the and fields spinor Th symplectic [4]. of spin [8], realization the e.g. of a see notions describe structure, the metaplectic of with counterparts manifolds symplectic are There Introduction ftecnrlgnrtro h eifco)a nidcn re inducing an as algebra factor) Lie Levi the to the rep associated of app modules Segal-Shale-Weil generator is central the which the of F-method, t of called components representation procedure metaplectic and the simple geometrical review basic projec briefly its homogeneous we the describe 3 Section and in two introduce we aim, this To eeaie em oue n iglrvcos4 dimension in 4 structure projective homogeneous The vectors singular 3 and modules Verma Generalized operator Dirac 2 symplectic the of Symmetries 1 Introduction Contents esalbifl omn ntecneto u ril.I Sect In article. our of content the on comment briefly shall We SL(3 f rjciestructure, Projective ooeeu oe ftedul oe fpoetv emtyin geometry projective of cover double dimensions. the real two of in the model in structure homogeneous operator projective Dirac homogeneous symplectic the of) and fields spinor plectic SL(3 Keywords: ls ypetcDrcoperator, Dirac symplectic ules, 00MteaisSbetClassification: Subject Mathematics 2010 f , nprdb h eut nsmere ftesmlci ia op Dirac symplectic the of symmetries on results the by Inspired R , ) R sl ) n h ypetcDrcoperator Dirac symplectic the and (3 . , R ) rjciesrcue ea-hl-elrepresentation, Segal-Shale-Weil structure, Projective sa raiigpicpefrtesmer ler fteso the of algebra symmetry the for principle organizing an as ai oíoá io řža erSomberg Petr Křižka, Libor Holíková, Marie SL(3 SL(3 f sl (3 , R operator , ) R . Abstract ) , n t aia aaoi uagba nti way this In subalgebra. parabolic maximal its and R ) 33,5D5 81R25. 53D05, 53C30, n h ypetcDirac symplectic the and i ftepeetsotatcei to is article short present the of aim e d perdi 1,btismeaning its but [1], in appeared ady eetto titdb character a by (twisted resentation ertclpoete.I eto 2 Section In properties. heoretical ftesmlci ia prtrin operator Dirac symplectic the of petcDrcoperator Dirac mplectic iesrcuei h eldimension real the in structure tive rsnainfrgnrlzdVerma generalized for presentation rfil n h ia prtron operator Dirac the and field or iesrcuei h eldimension real the in structure tive rmwr f(h obecover double (the of framework 2 ua xlnto ntrso the of terms in explanation tural idi h atScin4t the to 4 Section last the in lied trslso ymtiso the of symmetries on results nt h ymtygopo the of group symmetry The o esatwt motivation a with start we 1 ion eeaie em mod- Verma generalized rtr eraiesym- realize we erator, w eldmnin is dimensions real two uinsaeof space lution D s nthe in D s 7 5 2 1 . we produce the symplectic Dirac operator as an SL(3, R)-equivariant differential operator on the double covering of the real projective space RP2. f Let us highlight the meaning of the solution space for the symplectic Dirac operator Ds, regarded as an equivariant differential operator acting on the SL(3, R)-principal series representation induced from the character twisted metaplectic representation of SL(2, R). The solution space of f D on the symplectic space R2 was already determined in [2], as a consequence of the metaplectic s f Howe duality for the pair (mp(2, R), sl(2, R)). The underlying Harish-Chandra module of ker Ds for the Harish-Chandra pair (sl(3, R), SU(2)) with the maximal compact subgroup SU(2) ⊂ SL(3, R) is a unitarizable irreducible representation, equivalent to the exceptional representation of SL(3, R) f which is associated with the minimal coadjoint orbit, cf. [10], [12]. f As for the notation used throughout the article, we consider the pair (G, P ) consisting of a connected real reductive Lie group G and its parabolic subgroup P . In the Levi decomposition P = LU, L denotes the Levi subgroup and U the unipotent subgroup of P . We write g(R), p(R), l(R), u(R) for the real Lie algebras and g, p, l, u for the complexified Lie algebras of G, P , L, U, respectively. The symbol U applied to a Lie algebra denotes its universal enveloping algebra, and similarly applied to a Lie group denotes its double cover. f 1 Symmetries of the symplectic Dirac operator

In the present section we start with a motivation for our further considerations. Let (R2,ω) be the 2-dimensional symplectic vector space with the canonical symplectic form ω = dx ∧ dy, where 2 (x, y) are the canonical linear coordinate functions on R . We denote by ∂x, ∂y the coordinate vector ﬁelds and by e1,e2 corresponding symplectic frame ﬁelds acting on a symplectic spinor ϕ ∈ C[R2] ⊗C S(R) by

e1 · ϕ = iqϕ, e2 · ϕ = ∂qϕ. (1.1)

Here S(R) is the Schwartz space of rapidly decreasing complex functions on R equipped with the coordinate function q. The basis elements {X,Y,H} of the metaplectic Lie algebra mp(2, R), the double cover of the symplectic Lie algebra sp(2, R) ≃ sl(2, R), act on the function space C[R2] ⊗C S(R) by

i 2 1 i 2 X = −y∂x − 2 q , H = −x∂x + y∂y + q∂q + 2 , Y = −x∂y − 2 ∂q (1.2) and satisfy the commutation relations

[H,X]=2X, [X, Y ]= H, [H, Y ]= −2Y. (1.3)

The operators (1.2) preserve the homogeneity in the variables x, y, and the mp(2, R)-equivariant operators

1 Xs = y∂q + ixq, E = x∂x + y∂y + 2 ,Ds = iq∂y − ∂x∂q (1.4) form a Lie algebra isomorphic to sl(2, R). The operator Ds is termed the symplectic Dirac operator on R2. A differential operator P is called a symmetry of Ds provided there exists another differential ′ ′ operator P such that P Ds = DsP . Consequently, the symmetry differential operators preserve the solution space of Ds. The vector space of first order (in all variables x, y, q) symmetries was described in [1], and there is a Lie algebra structure given by the commutator of two symmetry differential operators. The commutators of elements in (1.2) with the symmetry differential operators classified in [1] yield the following result, whose proof is a straightforward but tedious computation. Lemma 1.1. The solution space of the symplectic Dirac operator on R2 is preserved by the following differential operators:

2 1) The couple of commuting diﬀerential operators

2 1 i 2 1 1 1 1 O1 = x ∂x + xy∂y − 2 xq∂q + 2 y∂q + 2 x = − 2 xH − yY + 2 xE + 2 x, (1.5) 2 1 i 2 1 1 1 O2 = xy∂x + y ∂y + 2 yq∂q + 2 xq + y = 2 yH − xX + 2 yE + 2 y (1.6) satisﬁes 3 3 [Ds, O1]= 2 xDs and [Ds, O2]= 2 yDs, (1.7) 3 3 which is equivalent to DsO1 = O1 + 2 x Ds and DsO2 = O2 + 2 y Ds, respectively. The operators O1, O2 increase the homogeneity in the variables x, y by one. 2) The couple of commuting diﬀerential operators

∂x and ∂y (1.8)

commutes with Ds and decreases the homogeneity in the variables x, y by one.

The operators O1, O2 turn out to be useful in the light of the following observation, whose proof is again straightforward and left to the reader. Theorem 1.2. The ﬁrst order symmetry diﬀerential operators

{∂x, ∂y,H,X,Y,E, O1, O2} (1.9) of the symplectic Dirac operator Ds fulﬁll the following non-trivial commutation relations

[∂x, O2]= −X, [∂y, O1]= −Y, 1 1 [∂y, O2]= 2 (3E + H), [∂x, O1]= 2 (3E − H), [O2,H]= −O2, [O1,H]= O1, [O2, E]= −O2, [O1, E]= −O1, [O , Y ]= O , [O ,X]= O , 2 1 1 2 (1.10) [∂x,H]= −∂x, [∂y,H]= ∂y, [∂x, Y ]= −∂y, [∂y,X]= −∂x, [∂x, E]= ∂x, [∂y, E]= ∂y, [H,X]=2X, [H, Y ]= −2Y, [X, Y ]= H.

The homomorphism of Lie algebras

h∂x, ∂y,H,X,Y,E, O1, O2i → sl(3, R) (1.11) given by

0 0 0 0 0 0 −∂x 7→ 1 0 0, −∂y 7→ 0 0 0, 0 0 0 1 0 0     0 1 0 0 0 1 O1 7→ 0 0 0, O2 7→ 0 0 0, 0 0 0 0 0 0     (1.12) 0 0 0 00 0 0 0 0 X 7→ 0 0 1, H 7→ 01 0 , Y 7→ 0 0 0, 0 0 0 0 0 −1 0 1 0       2 3 0 0 1 E 7→ 0 − 3 0  0 0 − 1  3  is an isomorphism of Lie algebras.

3 As will be proved in [11] by the techniques of tractor calculus, the operators (1.9) in the variables x, y, q are in fact all ﬁrst order symmetry diﬀerential operators of Ds in the base variables x, y. In the remaining part of our article we interpret the results of the present section in the framework of (the double cover of) the projective structure in real dimension 2.

2 Generalized Verma modules and singular vectors

It is well-known that the G-equivariant diﬀerential operators acting on principal series representations for G can be recognized in the study of homomorphisms between generalized Verma modules for the Lie algebra g. The latter homomorphisms are determined by the image of the highest weight vectors, referred to as the singular vectors and characterized as being annihilated by the positive nilradical u. An approach to ﬁnd precise positions of singular vectors in the representation space can be found in [7], [5], [6]. Let V denote a complex simple highest weight L-module, extended to P - module by U acting trivially. We denote by V∗ the (restricted) dual P -module to V. Any character λ ∈ HomP (p, C) yields a 1-dimensional representation Cλ of p by Xv = λ(X)v, X ∈ p, v ∈ C. (2.1)

λ Moreover, assuming that λ ∈ HomP (p, C) deﬁnes a group character e : P → GL(1, C) of P and denoting by ρ ∈ HomP (p, C) the character

1 ρ(X)= 2 tru ad(X), X ∈ p, (2.2) we may introduce a twisted P -module Vλ+ρ ≃ V ⊗C Cλ+ρ (with a twist λ + ρ) where p ∈ P acts λ+ρ on v ∈ Vλ+ρ ≃ V (the isomorphism of vector spaces) by e (p)p.v. In the rest of our article, V = S is one of the simple metaplectic submodules of the Segal-Shale-Weil representation twisted by character of GL(1, R)+. By abuse of notation, we call the highest weight modules induced from infinite-dimensional simple L-modules generalized Verma modules, though they are not the objects of the parabolic BGG category Op because of the lack of L-finiteness condition. However, most of structural results required in the present article carry over to this class of modules, cf. [6]. G V In general, for a chosen principal series representation of G on the vector space IndP ( λ+ρ) of smooth sections of the homogeneous vector bundle G ×P Vλ+ρ → G/P associated to a P -module Vλ+ρ, we compute the infinitesimal action

πλ : g → D(Ue) ⊗C End Vλ+ρ. (2.3)

Here D(Ue) denotes the C-algebra of smooth complex linear diﬀerential operators on Ue = UP ⊂ G/P (U is the Lie group whose Lie algebra is the opposite nilradical u(R) to u(R)), on the vector ∞ space C (Ue) ⊗CVλ+ρ of Vλ+ρ-valued smooth functions on Ue in the non-compact picture of the induced representation. ′ V V The dual vector space Do(Ue) ⊗C λ+ρ of λ+ρ-valued distributions on Ue supported on the unit coset o = eP ∈ G/P is D(Ue)⊗C End Vλ+ρ-module, and there is an U(g)-module isomorphism g V V ′ V Ag V Φλ : Mp ( λ−ρ) ≡ U(g) ⊗U(p) λ−ρ → Do(Ue) ⊗C λ+ρ ≃ u/Ie ⊗C λ+ρ. (2.4)

The exponential map allows to identify Ue with the nilpotent Lie algebra u(R). If we denote by Ag ′ u the Weyl algebra of the complex vector space u, then the vector space Do(Ue) can be identiﬁed Ag Ag as an u-module with the quotient of u by the left ideal Ie generated by all polynomials on u vanishing at the origin. Let (x1, x2,...,xn) be the linear coordinate functions on u and (y1,y2,...,yn) be the dual linear coordinate functions on u∗. Then the algebraic Fourier transform Ag Ag F : u → u∗ (2.5)

4 is given by

F(xi)= −∂yi , F(∂xi )= yi (2.6) for i =1, 2,...,n, and leads to a vector space isomorphism

g ∗ ∼ g ∗ τ : A /Ie ≃ C[u ] −→ A ∗ /F(Ie) ≃ C[u ], u u (2.7) Q mod Ie 7→ F(Q) mod F(Ie) Ag for Q ∈ u. The composition of (2.4) and (2.7) gives the vector space isomorphism V ∼ ′ V ∼ C ∗ V τ ◦ Φλ : U(g) ⊗U(p) λ−ρ −→Do(Ue) ⊗C λ+ρ −→ [u ] ⊗C λ+ρ, (2.8) ∗ thereby inducing the g-module action πˆλ on C[u ] ⊗C Vλ+ρ. Definition 2.1. Let V be a complex simple highest weight L-module, extended to a P -module by U acting trivially. We define the L-module g V u g V Mp ( ) = {v ∈ Mp ( ); Xv =0 for all X ∈ u}, (2.9) which we call the vector space of singular vectors. The vector space of singular vectors is for an infinite-dimensional complex simple highest weight g u P -module V an infinite-dimensional L-module. In the case when Mp (V) is a completely reducible L-module, we denote by W one of its simple L-submodule. Then we obtain the U(g)-module g g homomorphism from Mp (W) to Mp (V), and moreover we have g W g V W g V u Hom(g,P )(Mp ( ),Mp ( )) ≃ HomL( ,Mp ( ) ). (2.10)

We introduce the L-module

∗ F ∗ Sol(g, p; C[u ] ⊗C Vλ+ρ) = {f ∈ C[u ] ⊗C Vλ+ρ;π ˆλ(X)f =0 for all X ∈ u}, (2.11) and by (2.8), there is an L-equivariant isomorphism g V u ∼ C ∗ V F τ ◦ Φλ : Mp ( λ−ρ) −→ Sol(g, p; [u ] ⊗C λ+ρ) . (2.12) ∗ The action of πˆλ(X) on C[u ] ⊗C Vλ+ρ produces a system of partial differential equations for the ∗ F elements in Sol(g, p; C[u ] ⊗CVλ+ρ) , which makes it possible to describe its structure completely in particular cases of interest as the solution space of the systems of partial differential equations. The formulation above has the following classical dual statement (cf. [3] for the standard formulation in the category of finite dimensional inducing P -modules, or [6] for its extension to inducing modules with infinitesimal character), which explains the relationship between the geometrical problem of finding G-equivariant differential operators between induced representations and the algebraic problem of finding homomorphisms between generalized Verma modules. Let V and W be two simple highest weight P -modules. Then the vector space of G-equivariant differential oper- G V G W ators HomDiff(G)(IndP ( ), IndP ( )) is isomorphic to the vector space of (g, P )-homomorphisms g ∗ g ∗ Hom(g,P )(Mp (W ),Mp (V )).

3 The homogeneous projective structure in dimension 2

A projective structure on a smooth manifold M of real dimension n ≥ 2 is a class [∇] of projectively equivalent torsion-free connections, which define the same family of unparametrized geodesics. A connection is projectively flat if and only if it is locally equivalent to a flat connection. Given any non-vanishing volume form ω on M, there exists a unique connection in the projective class such that ∇ω = 0. In the case n > 2 (n = 2), the vanishing of the Weyl curvature tensor (the Cotton curvature tensor) for ∇ is equivalent to the projective flatness of [∇] and consequently,

5 the existence of a local isomorphism with the flat model of n-dimensional projective geometry on RPn equipped with the flat projective structure given by the absolute parallelism. The homogeneous (flat) model of projective geometry in the real dimension 2 is RP2 ≃ G/P , where G is the connected simple real Lie group SL(3, R) and P ⊂ G the parabolic subgroup stabilizing the line [v] ∈ R3 generated by a non-zero vector v in the defining representation R3 of G. Although our construction of equivariant differential operators is local, the passage to the generalized flag manifold and sections of associated vector bundles induced from half inte- gral modules (e.g., the simple metaplectic submodules of the Segal-Shale-Weil representation) on G/P requires the double (universal) cover G = SL(3, R) and its parabolic subgroup P . The Lie group SL(3, R) acts transitively on S2 ≃ CP1, the double (universal) cover of RP2, with e f e parabolic stabilizer P = (GL(1, R) × SL(2, R)) ⋉R2. We notice that the double (universal) cover f + SL(3, R)/P ≃ S2 ≃ CP1 is a symplectic manifold, while RP2 is non-orientable and hence not e f symplectic. f e The questions discussed in our article can be treated by infinitesimal methods, and so we introduce the complexified Lie algebra g = sl(3, C) of G and the Cartan subalgebra h ⊂ g by

h = {diag(a1,a2,a3); a1 + a2 + a3 =0, a1,a2,a3 ∈ C}. (3.1) ∗ For i =1, 2, 3, we deﬁne εi ∈ h by εi(diag(a1,a2,a3)) = ai. The root system of g with respect to + h is ∆ = {±(εi − εj ); 1 ≤ i < j ≤ 3}, the positive root system is ∆ = {εi − εj ; 1 ≤ i < j ≤ 3} with the subset of simple roots Π = {α1, α2}, α1 = ε1 − ε2, α2 = ε2 − ε3, and the fundamental ∗ weights are ω1 = ε1, ω2 = ε1 +ε2. The subset Σ= {α2} of Π generates a root subsystem ∆Σ ⊂ h , and we associate to Σ the standard parabolic subalgebra p of g with p = l ⊕ u. The reductive Levi subalgebra l of p is

l = h ⊕ gα, (3.2)

αM∈∆Σ and the nilradical u of p and the opposite nilradical u are

u = gα, u = g−α, (3.3) ∈ M+r + ∈ M+r + α ∆ ∆Σ α ∆ ∆Σ respectively. The Σ-height htΣ(α) of a root α ∈ ∆ is deﬁned by

htΣ(a1α1 + a2α2)= a1, (3.4) g g g and is a |1|-graded Lie algebra with respect to the grading given by i = α∈∆, htΣ(α)=i α for Z CL2 C2 0 6= i ∈ , and g0 = h ⊕ α∈∆, htΣ(α)=0gα. In particular, we have u = g1 ≃ , u = g−1 ≃ and l = g0 ≃ C ⊕ sl(2, C). L The basis {e1,e2} of the root spaces in the nilradical u is given by 0 1 0 0 0 1 e1 = 0 0 0, e2 = 0 0 0, (3.5) 0 0 0 0 0 0     the basis {f1,f2} of the root spaces in the opposite nilradical u is 0 0 0 0 0 0 f1 = 1 0 0, f2 = 0 0 0, (3.6) 0 0 0 1 0 0     and ﬁnally the basis {h0,h,e,f} of the Levi subalgebra l is 0 0 0 00 0 0 0 0 e = 0 0 1, h = 01 0 , f = 0 0 0, 0 0 0 0 0 −1 0 1 0       (3.7) 1 0 0 1 h0 = 0 − 2 0 , 0 0 − 1  2 

6 where h0 generates a basis of the center z(l) of l. Any character χ ∈ HomP (p, C) is given by

χ = λω1, λ ∈ C, (3.8) C ∗ e where ω1 ∈ HomP (p, ) is equal to ω1 ∈ h regarded as trivially extended to p = h⊕gα2 ⊕g−α2 ⊕u. Throughout the article we use the simpliﬁed notation λ for λω1. The vector ρ ∈ HomP (p, C) deﬁnede by the formula (2.2) is then e 3 ρ = 2 ω1. (3.9) e 4 SL(3, R) and the symplectic Dirac operator In thisf section we retain the notation in Section 3 and describe the class of representations of g on the space of sections of vector bundles on G/P associated to the simple metaplectic submodules of the Segal-Shale-Weil representation S of P twisted by characters λ + ρ ∈ Hom e(p, C). λ+ρe e P The induced representations in question are described in the non-compact picture, given by e restricting sections to the open Schubert cell Ue ⊂ G/P which is isomorphic by the exponential map to the opposite nilradical u(R). We denote by (ˆx, yˆ) the linear coordinate functions on u(R) e e with respect to the basis {f1,f2} of u(R), and by (x, y) the dual linear coordinate functions on ∗ R Ag u ( ). The Weyl algebra u is generated by

{x,ˆ y,ˆ ∂xˆ, ∂yˆ} (4.1) Ag and the Weyl algebra u∗ by

{x, y, ∂x, ∂y}. (4.2)

For a p-module (σ, V), σ : p → gl(V), the twisted p-module (σλ, Vλ), σλ : p → gl(Vλ), with a C twist λ ∈ HomPe(p, ) is deﬁned by

σλ(X)v = σ(X)v + λ(X)v (4.3) for all X ∈ p and v ∈ Vλ ≃ V (the isomorphism of vector spaces). We use the following realization of the simple mp(2, C)-submodules of the Segal-Shale-Weil representation. The Fock model is the unitarizable mp(2, C)-module S = C[q]. Since the simple part of the Levi algebra l is ls ≃ mp(2, C), we realize the simple metaplectic submodules of the Segal-Shale-Weil representation as the representations of ls on the subspace of polynomials of even and odd degree, respectively. The generators act as

i 2 1 i 2 σ(e)= 2 ∂q , σ(h)= −q∂q − 2 , σ(f)= 2 q . (4.4)

The scalar product h· , ·i: S ⊗C S → C on S is deﬁned through the ls-equivariant embedding into the space of Schwartz functions ι: S → S(R),

hp1,p2i = ι(p1)ι(p2) dq for all p1,p2 ∈ S. (4.5) ZR

The representation of ls is then extended to a representation of p by the trivial action of the center z(l) of l and by the trivial action of the nilradical u of p. We retain the same notation σ : p → gl(S) for the extended action of the parabolic subalgebra p of g. In what follows, we are interested in S C the twisted p-module σλ : p → gl( λ) with a twist λ ∈ HomPe(p, ). C Ag S Ag Theorem 4.1. Let λ ∈ HomPe(p, ). Then the embedding of g into u ⊗C End λ+ρ and u∗ ⊗C End Sλ+ρ is given by

7 1)

πλ(f1)= −∂xˆ, πλ(f2)= −∂yˆ, (4.6)

πˆλ(f1)= −x, πˆλ(f2)= −y; (4.7)

i 2 1 i 2 πλ(e)= −y∂ˆ xˆ + 2 ∂q , πλ(h)= −x∂ˆ xˆ +ˆy∂yˆ − q∂q − 2 , πλ(f)= −x∂ˆ yˆ + 2 q , 3 3 (4.8) πλ(h0)= 2 (ˆx∂xˆ +ˆy∂yˆ)+ λ + 2 ,

i 2 1 i 2 πˆλ(e)= x∂y + 2 ∂q , πˆλ(h)= x∂x − y∂y − q∂q − 2 , πˆλ(f)= y∂x + 2 q , 3 3 (4.9) πˆλ(h0)= − 2 (x∂x + y∂y)+ λ − 2 ;

3 1 1 i 2 πλ(e1)=ˆx(ˆx∂xˆ +ˆy∂yˆ + λ + 2 )+ 2 xˆ q∂q + 2 − 2 yqˆ , 3 1 1 i 2 (4.10) πλ(e2)=ˆy(ˆx∂xˆ +ˆy∂yˆ + λ + 2 ) − 2 yˆ q∂q + 2 − 2 x∂ˆ q ,

1 1 πˆλ(e1)= ∂x x∂x + y∂y − λ + 4 + 2 q(iq∂y − ∂x∂q), 1 i (4.11) πˆλ(e2)= ∂y x∂x + y∂y − λ + 4 − 2 ∂q(iq∂y − ∂x∂q). Proof. The proof of this claim for the trivial representation of p instead of the simple metaplectic submodules of the Segal-Shale-Weil representation follows from Theorem 1.3 in [5]. For the Segal- Shale-Weil representation it follows by a straightforward veriﬁcation of all commutation relations for the Lie algebra g. ∗ F 3 C C S Theorem 4.2. The space Sol(g, p; [u ] ⊗ λ+ρ) is for λ = 4 ω1 non-trivial and contains the L-submodule XsSλ+ρ. e Proof. By Theorem (4.1), the sl(2, C)-algebra structure for the operators (1.4) implies

1 1 5 i πˆλ(e1)Xsv0 = ∂x 1 − λ + 4 Xsv0 + 2 q[Ds,Xs]v0 = 4 − λ [∂x,Xs]v0 − 2 qv0 5 1 = 4 − λ − 2 iqv0 =0, 1 1 5 1 πˆλ(e2)Xsv0 = ∂y 1 − λ + 4 Xsv0 − 2 i∂q[Ds,Xs]v0 = 4 − λ [∂y,Xs]v0 − 2 ∂qv0 5 1 = 4 − λ − 2 ∂qv0 =0 S 3 for all v0 ∈ λ+ρ, provided λ = 4 . The proof is complete. k Remark. We notice that the vectors Xs v for k > 1 are not singular vectors. For example, a straightforward computation reveals

k 1 k−1 k(k−1) k−2 k(k−1) k−1 πˆλ(e1)Xs v0 = k − λ + 4 ikqXs + i 2 yXs v − i 4 qXs v, (4.12) which is non-zero for all λ ∈ C. It is not difficult to exploit the results in [2] in order to classify the complete set of solutions of the system in Theorem 4.1, but the detailed analysis goes beyond the scope of our article. Theorem 4.2 has the following classical corollary, which explains the relationship between the geometrical problem of finding G-equivariant differential operators between induced representations and the algebraic problem of finding homomorphisms between generalized Verma modules, 2 cf. [3], [6]. There is a double cover P = (GL(1, R)+ × SL(2, R)) ⋉R of the maximal parabolic subgroup P of the Lie group SL(3, R), which splits over the unipotent subgroup N ≃ R2 in the e f

8 Langlands-Iwasawa decomposition of P , [13]. Let us note that the extension cocycle splits over the ﬁeld of complex numbers. e 2 Theorem 4.3. Let G = SL(3, R) and let P = (GL(1, R)+ × SL(2, R)) ⋉R be the maximal parabolic subgroup of G, whose unipotent subgroup in the Langlands-Iwasawa decomposition of e f e f P is N ≃ R2. For V = S we have V∗ ≃ S∗ . Then the singular vector constructed in Theorem e λ −λ 4.2 corresponds to the G-equivariant diﬀerential operator, given in the non-compact picture of the e induced representations by e ∞ R S∗ ∞ R S∗ Ds : C (u( ), 3 ωe ) →C (u( ), 9 ωe ), 4 1 4 1 (4.13) ϕ 7→ (iq∂yˆ − ∂xˆ∂q)ϕ.

The inﬁnitesimal intertwining property of Ds is

∗ ∗ Dsπ− 3 (X)= π 3 (X)Ds (4.14) 4 4 for all X ∈ sl(3, R). With abuse of notation we used the same symbol Ds and the same terminology ”the symplectic Dirac operator” as in (1.4) due to the coincidence of (1.4) and (4.13). Let us ﬁnally explain the notion of the dual representation S∗. Let us deﬁne a non-degenerate pairing (· , ·): S ⊗C S → C on S by the formula

(p1(q),p2(q)) = p1(∂q)p2(q)|q=0. (4.15)

Then we can identify the (restricted) dual space to S with S and the structure of the dual p-module on S is given as follows: the generators of mp(2, C) act on S∗ by

∗ i 2 ∗ 1 ∗ i 2 σ (e)= − 2 q , σ (h)= q∂q + 2 , σ (f)= − 2 ∂q , (4.16) while the generator h0 of the center z(l) ⊂ l and of the nilradical u of p act trivially. We notice that this representation is compatible with (1.2).

Acknowledgement

The authors gratefully acknowledge the support of the grant GACR P201/12/G028 and SVV- 2016-260336.

References

[1] Hendrik De Bie, Marie Holíková, and Petr Somberg, Basic aspects of symplectic Cliﬀord analysis for the symplectic Dirac operator, arXiv:1511.04189 (2015). [2] Hendrik De Bie, Petr Somberg, and Vladimír Souček, The metaplectic Howe duality and polynomial solutions for the symplectic Dirac operator, J. Geom. Phys. 75 (2014), 120–128. [3] David H. Collingwood and Brad Shelton, A duality theorem for extensions of induced highest weight modules, Paciﬁc J. Math. 146 (1990), no. 2, 227–237. [4] Katharina Habermann and Lutz Habermann, Introduction to symplectic Dirac operators, Lecture Notes in Mathematics, vol. 1887, Springer-Verlag, Berlin, 2006. [5] Libor Křižka and Petr Somberg, Algebraic analysis on scalar generalized Verma modules of Heisenberg parabolic type I.: An-series, arXiv:1502.07095 (2015). [6] , Algebraic analysis on scalar generalized Verma modules of Heisenberg parabolic type II.: Cn,Dn-series, (in preparation).

9 [7] Toshiyuki Kobayashi, Bent Ørsted, Petr Somberg, and Vladimír Souček, Branching laws for Verma modules and applications in parabolic geometry. I, Adv. Math. 285 (2015), 1796–1852. [8] Bertram Kostant, Symplectic spinors, Symposia Mathematica, Progress in Mathematics, vol. XIV, Academic Press, London, 1974, pp. 139–152. [9] , Verma modules and the existence of quasi-invariant diﬀerential operators, Non- Commutative Harmonic Analysis, Lecture Notes in Mathematics, vol. 466, Springer, Berlin, 1975, pp. 101–128. [10] Bent Ørsted, Generalized gradients and Poisson transforms, Global analysis and harmonic analysis, Semin. Congr., vol. 4, Soc. Math. France, Paris, 2000, pp. 235–249.

[11] Petr Somberg and Josef Šilhan, Higher symmetries of the symplectic Dirac operator, (in preparation). [12] Pierre Torasso, Quantiﬁcation géométrique, opérateurs d’entrelacement et représentations unitaires de SL3(R), Acta Math. 150 (1983), no. 1, 153–242. [13] Joseph A. Wolf,f Unitary representations of maximal parabolic subgroups of the classical groups, vol. 8, Mem. Amer. Math. Soc., no. 180, American Mathematical Society, Providence, 1976.

(M. Holíková) Department of Mathematics and Mathematical Education, Faculty of Education, Charles University, Magdalény Rettigové 4, 116 39 Praha 1, Czech Republic E-mail address: [email protected] (L. Křižka) Mathematical Institute of Charles University, Sokolovská 83, 18000 Praha 8, Czech Republic E-mail address: [email protected] (P. Somberg) Mathematical Institute of Charles University, Sokolovská 83, 18000 Praha 8, Czech Republic E-mail address: [email protected]